Table 2 Possible Study Designs for Rare Disease Research

Experimental

Crossover RCT

• Patients are guaranteed to receive active treatment and spend less time on placebo

• Patients receive two interventions in sequence randomly, with a washout period between interventions

• Each participant serves as his or her own control, thereby reducing sample size requirements

• Latin square allows for multiple interventions in randomized sequence

• Each intervention appears only once in each sequence and intervention period

• Design variations include N-of-1 trials

Adaptive RCT

• This design can increase the proportion of patients assigned to the more favorable treatment, which can contribute to a greater number of willing participant

• Adaptive treatment allocation designs test the null hypothesis in a series of interim analyses; these analyses then influence subsequent randomization in the next phase

• Bayesian analyses (allowing updates of prior probabilities) or frequentist approaches can be used

• Adaptive treatment allocation designs allow the probability of being randomized to an intervention to change during the enrollment period; the probability of being randomized will increasingly favor the arm with the more promising results (play the winner) or increasingly penalize the arm with less promising results (drop the loser

• Adaptive designs can be used to narrow from a selection of doses (ranking and selection designs) rather than rejecting a null hypothesis

• Adaptive designs can be used to select among subpopulations and thereby balance covariates (covariate-adaptive randomization) and help address underlying heterogeneity

• Design variations include sequential RCTs and ranking and selection RCTs

Randomized Placebo Phase

• This design minimizes the length of time that patients are on placebo with all patients receiving treatment in the end. Limited time on placebo is beneficial for conditions with a rapid unfavorable evolution

• Design variations include stepped wedge, randomized withdrawal, and three-stage trials. In stepped wedged designs, randomization occurs at crossover to different treatment

Risk-based Allocation

• Low-risk patients are randomized to high-dose and standard treatment, high-risk patients receive high-dosetreatment, thereby addressing concerns about the ethics of withholding treatment from high-risk patients

• A combined analysis allows the prediction of the added benefit of high-dose treatment

Non-experimental

Case-control study^a

• This design offers patients with diseases that have long latency periods the opportunity to participate in research

• Study participants with the disease (cases) are compared to participants without the disease (controls) in an effort to identify factors that may contribute to a particular outcome

• To address concern that cases and controls may differ in characteristics other than the condition studied, cases and controls can be matched on other characteristics (ex. Age, race, sex)

Cohorts with historic controls [Natural History Studies]

• This design provides patients with the opportunity to learn about different treatments

• Comparison of prospectively treated patients with historic controls reduces recruitment burden for the control arm

Pre-post designs

• Patients receive usual care or standard intervention followed by tested treatment. Patients receive an active treatment throughout the course of the study

• Requires a detailed understanding of the natural history of the disease to avoid issues of regression to the mean